Dynamic Characterization of Tilting Pad Journal Bearings From Component and System Level Testing
Conference Paper
Overview
Identity
Additional Document Info
Other
View All
Overview
abstract
The dynamic response of a direct lube, 5-pad, rocker-back pivot tilting pad bearing is characterized in a controlled motion (component level) test rig, and in a spin bunker (full system level) using a dummy rotor mounted on two identical bearings. In the component level test, the force coefficients (stiffness, damping, mass) are identified from pseudorandom excitations using a 2-DOF model. N-DOF system including the pad motions has been shown to yield frequency dependent coefficients that warrant the use of asynchronous coefficients for stability analysis in centrifugal compressors. However, experimental results showed that the real part of the dynamic stiffness is well represented as a constant stiffness and mass coefficients while the imaginary part yields a constant damping coefficient (i.e. frequency independent).In the system level test, a dedicated dummy rotor (representative of a high speed centrifugal compressor rotor) is excited by a magnetic shaker throughout a frequency range covering the rotor modes of interest while spinning at constant speed. From the rotor harmonic response the damping of each mode is extracted using a curve-fitting method based on a 1-DOF model for a given set of speeds.The dummy rotor test provides reference values for system logarithmic decrement and further validates the component level test results. The logarithmic decrement prediction using identified bearing force coefficients are in good agreement with the experimental results. In addition, using for prediction identified coefficients in a classical K-C-M or synchronous K-C form yields similar results (within 15%). This indicates that for the given bearing geometry (clearance, offset and size) and operating conditions, synchronously reduced force coefficients are adequate for stability analysis. Comparison of the identified force coefficients with results from commercially available code yielded reasonable agreement on direct coefficients while some discrepancies are highlighted on the cross-coupled coefficients.
